At the bottom of circuit_predict_and_launch_new() we check ``` /* Finally, check to see if we still need more circuits to learn * a good build timeout. But if we're close to our max number we * want, don't do another -- we want to leave a few slots open so * we can still build circuits preemptively as needed. */ if (num < MAX_UNUSED_OPEN_CIRCUITS-2 && get_options()->LearnCircuitBuildTimeout && circuit_build_times_needs_circuits_now(&circ_times)) { flags = CIRCLAUNCH_NEED_CAPACITY; log_info(LD_CIRC, "Have %d clean circs need another buildtime test circ.", num); circuit_launch(CIRCUIT_PURPOSE_C_GENERAL, flags); return; } ``` Here we check get_options()->LearnCircuitBuildTimeout directly, rather than calling circuit_build_times_disabled(). That means we do these exploratory circuits even when cbt is disabled by consensus, or when we're a directory authority, or when we've failed to write cbt history to our state file lately. We also have the same bug in circuit_expire_old_circuits_clientside() where ``` if (get_options()->LearnCircuitBuildTimeout && circuit_build_times_needs_circuits(&circ_times)) { /* Circuits should be shorter lived if we need more of them * for learning a good build timeout */ cutoff.tv_sec -= IDLE_TIMEOUT_WHILE_LEARNING; } else { cutoff.tv_sec -= get_options()->CircuitIdleTimeout; } ```